Information processing device, information processing method, and information processing program

Abstract

The present invention provides an information processing device, an information processing method, and an information processing program that can appropriately generate a control model that predicts the optimal operation of a controlled object. [Solution] The information processing device 10 includes a causal relationship information generation unit 14a that generates sensor causal relationship information 12e; an influence on controlled variables calculation unit 14b that calculates the degree of influence on controlled variables, indicating the degree of influence of changes to the sensor measurement data 12b on controlled variables, based on sensor measurement data 12b, controlled variable information 12c, and causal relationship information 12e; an influence from control variables calculation unit 14c that calculates the degree of influence from control variables, indicating the degree of influence of changes to the control variables on sensor measurement data 12b, based on sensor measurement data 12b, control variable information 12d, and causal relationship information 12e; and a selection unit 14d that selects input variables for the control model 12i based on the degree of influence on controlled variables and the degree of influence from control variables.

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing method, and an information processing program, and more particularly to an information processing device, an information processing method, and an information processing program suitable for generating control models. [Background technology]

[0002] Regarding the generation of predictive models, for example, the technology described in Patent Document 1 is known. Specifically, Patent Document 1 discloses a calculation unit that calculates an index indicating the controllability of a plant for each of the multiple explanatory variables, based on data of the target variable to be predicted and data of multiple explanatory variables used to explain the target variable, from data measured at the plant, and a generation unit that generates a predictive model that predicts the target variable by machine learning, based on the data of the explanatory variables selected based on the index calculated by the calculation unit. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2021-174125 [Overview of the project] [Problems that the invention aims to solve]

[0004] The technology described in Patent Document 1 generates a predictive model with good controllability for the plant by generating data of multiple feature variables suitable for predicting the target variable from data of multiple explanatory variables that explain the target variable to be predicted, among the data measured at the plant. However, there is room for improvement in generating control models that predict the optimal operation of a controlled object using measurement data related to the controlled object as explanatory variables. For example, including items that are direct results of the operation predicted by the control model as explanatory variables in the control model may reduce the control performance of the control model.

[0005] Therefore, the present invention provides an information processing device, an information processing method, and an information processing program that can appropriately generate a control model that predicts the optimal operation of a controlled object. [Means for solving the problem]

[0006] To solve the above problems, the information processing device according to the present invention is characterized by comprising: a data acquisition unit that receives input of sensor connection relationship information, sensor measurement data, operation variable information of the controlled object, and controlled variable information of the controlled object; a causal relationship information generation unit that generates causal relationship information based on the sensor connection relationship information and sensor measurement data; an influence on controlled variables calculation unit that calculates the degree of influence on controlled variables, indicating the degree of influence of changes in sensor measurement data on controlled variables, based on the sensor measurement data, controlled variable information, and causal relationship information; an influence from operation variables calculation unit that calculates the degree of influence from operation variables, indicating the degree of influence of changes in operation variables on the sensor measurement data, based on the sensor measurement data, operation variable information, and causal relationship information; and a selection unit that selects input variables for a control model based on the degree of influence on controlled variables and the degree of influence from operation variables.

[0007] Furthermore, the present invention relates to an information processing method for an information processing apparatus comprising: a data acquisition unit; a causal relationship information generation unit; an influence degree calculation unit for the controlled variable; an influence degree calculation unit from an operator variable; and a selection unit, wherein the data acquisition unit receives input of sensor connection relationship information, sensor measurement data, operator variable information for the controlled object, and controlled variable information for the controlled object; the causal relationship information generation unit generates causal relationship information based on the sensor connection relationship information and sensor measurement data; the influence degree calculation unit for the controlled variable calculates an influence degree for the controlled variable indicating the degree of influence of a change in sensor measurement data on the controlled variable based on the sensor measurement data, controlled variable information, and causal relationship information; the influence degree calculation unit from an operator variable calculates an influence degree from an operator variable indicating the degree of influence of a change in an operator variable on the sensor measurement data based on the sensor measurement data, operator variable information, and causal relationship information; and the selection unit selects input variables for a control model based on the influence degree for the controlled variable and the influence degree from an operator variable.

[0008] Furthermore, the information processing program according to the present invention is a program for a processor to implement the following functions: a function to receive input of sensor connection relationship information, sensor measurement data, operation variable information of the controlled object, and controlled variable information of the controlled object; a function to generate causal relationship information based on the sensor connection relationship information and sensor measurement data; a function to calculate the degree of influence on the controlled variable, indicating the degree of influence of changes in sensor measurement data on the controlled variable, based on the sensor measurement data, controlled variable information, and causal relationship information; a function to calculate the degree of influence from the operation variable, indicating the degree of influence of changes in the operation variable on the sensor measurement data, based on the sensor measurement data, operation variable information, and causal relationship information; and a function to select input variables for a control model based on the degree of influence on the controlled variable and the degree of influence from the operation variable. [Effects of the Invention]

[0009] According to the present invention, it is possible to provide an information processing apparatus, an information processing method, and an information processing program that can appropriately generate a control model for predicting an optimal operation of a control target. Problems, configurations, and effects other than those described above will be clarified by the following description of embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] [Figure 1] FIG. 1 is an overall schematic configuration diagram including an information processing apparatus according to a first embodiment of the present invention, and is a functional block diagram of the information processing apparatus. [Figure 2] FIG. 2 is a diagram showing an example of the hardware configuration of the information processing apparatus according to the first embodiment of the present invention. [Figure 3] FIG. 3 is an explanatory diagram showing a specific example of a plant that is a control target of a control model generated by the information processing apparatus shown in FIG. 1. [Figure 4] FIG. 4 is an explanatory diagram of sensor connection relationship information in the information processing apparatus shown in FIG. 1. [Figure 5A] FIG. 5A is an explanatory diagram showing a specific example of a plant that is a control target of a control model generated by the information processing apparatus shown in FIG. 1, the plant including sensors connected via signals. [Figure 5B] FIG. 5B is an explanatory diagram of sensor connection relationship information of the plant shown in FIG. 5A. [Figure 6] FIG. 6 is a diagram showing an example of sensor measurement data of a sensor included in the plant shown in FIG. 3. [Figure 7] FIG. 7 is a diagram showing an example of causal relationship information of a sensor included in the plant shown in FIG. 3. [Figure 8] FIG. 8 is a flowchart showing the processing flow of the information processing apparatus shown in FIG. 1. [Figure 9A] FIG. 9 is an explanatory diagram of the degree of influence on controlled variables in the plant shown in FIG. 3. [Figure 9B] FIG. 10 is an explanatory diagram of the degree of influence from manipulated variables in the plant shown in FIG. 3. [Figure 10] FIG. 11 is a diagram showing an example of a degree-of-influence table for controlled variables in the plant shown in FIG. 3. [Figure 11]Figure 3 shows an example of an influence table from the operational variables in the plant shown. [Figure 12] Figure 3 shows an example of a control model input variable selection table for the plant shown in the diagram. [Figure 13] Figure 3 shows an example of displaying the selected sensors as input variables for the control model in the plant shown in Figure 3, overlaid on the plant configuration diagram. [Figure 14] Figure 3 shows an example of displaying the degree of influence of each sensor on the controlled variable and the degree of influence from the manipulated variable in the plant shown in Figure 3 on a two-dimensional plane. [Modes for carrying out the invention]

[0011] This specification describes, as an example, a case in which an information processing device 10 (see Figure 1) generates a control model for a predetermined plant 20 (see Figure 1) as the control target. Here, a "control model" is a model that predicts the optimal operation of the control target according to its state. Examples of the plant 20 as the control target include power plants, chemical plants, as well as oil refineries, steel plants, food processing plants, pharmaceutical plants, and water treatment plants. Note that the control target of the control model generated by the information processing device 10 (see Figure 1) is not limited to plants, but may also be robots, vehicles, ships, aircraft, etc. The following describes embodiments of the present invention with reference to the drawings. [Examples]

[0012] <Configuration of the information processing system> Figure 1 is a schematic overall configuration including the information processing device according to Embodiment 1 of the present invention, and is a functional block diagram of the information processing device. As shown in Figure 1, the information processing system 100 is a system that generates a control model used to control the equipment of the plant 20 using the information processing device 10. The information processing system 100 includes the information processing device 10, a sensor 30, an input device 40, a display device 50, and a cloud 60, and these are connected in a predetermined manner by wired or wireless connections. Below, we will briefly describe the plant 20, which is the target of the control model generated by the information processing device 10, as well as the sensor 30, input device 40, display device 50, etc., and then describe the information processing device 10 in detail.

[0013] Plant 20 is a facility equipped with operating terminals such as valves and heaters, as in a chemical plant. Sensor 30 acquires predetermined detection values ​​and environmental information from the equipment in Plant 20. As such sensors 30, cameras, distance sensors, radar, force sensors, temperature sensors, angle sensors, flow sensors, pressure sensors, voltage sensors, current sensors, etc., can be used as appropriate.

[0014] The sensor 30 may be installed inside the plant 20 or outside the plant 20. One example of a sensor 30 installed outside the plant 20 is an ambient temperature sensor that detects the temperature of the surrounding environment of the plant 20. Other devices that acquire data such as power demand related to the control of the plant 20 from outside the plant 20 are also included in the sensor 30. Other devices that are included in the sensor 30 include soft sensors calculated from the detected values ​​of the sensor 30 and indicators of the operation history of the plant 20 input by operator M1 via the input device 40. Here, a soft sensor includes devices that obtain detected values ​​with different attributes from the detected values ​​of the sensor 30, or devices that convert the detected values ​​of the sensor 30 into values ​​that have been processed, for example, by smoothing.

[0015] The input device 40 is used when the operator M1 inputs predetermined control information or commands to the information processing device 10. For example, a keyboard, mouse, or joystick can be used as such an input device 40.

[0016] The display device 50 is used to present the calculation results of the information processing device 10 to the operator M1. For example, a liquid crystal display or an organic EL display can be used as such a display device 50. Alternatively, a touch-panel mobile terminal that combines the functions of both an input device 40 and a display device 50, such as a smartphone or tablet, may be used.

[0017] Cloud 60 has a server (not shown) that performs predetermined communication with the input device 40, display device 50, and information processing device 10 via a network. The processing results of the server are provided to the information processing device 10 via the network and are also displayed on the display device 50 as appropriate.

[0018] <Configuration of the information processing device> The information processing device 10 is a device that generates a control model for the plant 20. A personal computer, tablet, or smartphone may be used as such an information processing device 10. Alternatively, the information processing device 10 may be configured by connecting multiple computers via communication lines or a network. For example, the functions of the information processing device 10 may be distributed and implemented across multiple computers such as cloud servers or edge servers.

[0019] As shown in Figure 1, the information processing device 10 comprises a data acquisition unit 11, a storage unit 12, a communication unit 13, and a calculation unit 14. These data acquisition unit 11, storage unit 12, communication unit 13, and calculation unit 14 are connected to each other via an internal bus (not shown) so that they can communicate with one another.

[0020] Figure 2 shows an example of the hardware configuration of an information processing device according to Embodiment 1 of the present invention. As shown in Figure 2, the information processing device 10 has a hardware configuration that includes a processor 10a, RAM 10b (Random Access Memory), ROM 10c (Read Only Memory), storage 10d such as a Hard Disk Drive or Solid State Drive, a communication interface 10e, an input / output interface 10g, and a media interface 10h, all of which are connected to each other via an internal bus 10i so that they can communicate with one another.

[0021] The processor 10a shown in Figure 2 may be, for example, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), or a GPGPU (General-purpose computing on graphics processing units). The processor 10a is the hardware of the arithmetic unit 14 (see Figure 1) of the information processing device 10. The RAM 10b, ROM 10c, and storage 10d are the hardware of the memory unit 12 (see Figure 1) of the information processing device 10. The processor 10a reads a predetermined program stored in the ROM 10c or storage 10d and loads it into the RAM 10b to execute a predetermined process.

[0022] The communication interface 10e shown in Figure 2 communicates with the equipment, sensors 30, and cloud 60 of the plant 20. The input / output interface 10g receives data from the input device 40 and outputs data to the display device 50. The communication interface 10e and the input / output interface 10g function as the communication unit 13 of the information processing device 10 (see Figure 1).

[0023] The media interface 10h is to which external recording media 70 such as an optical disc 70a and a USB memory 70b are appropriately connected, and functions as the data acquisition unit 11 of the information processing device 10 (see Figure 1). Such a media interface 10h could be an optical disc drive for playing the optical disc 70a or a USB interface for reading the USB memory 70b. Note that "USB" is a registered trademark.

[0024] Figure 3 is an explanatory diagram showing a specific example of a plant that is the target of the control model generated by the information processing device shown in Figure 1. In the example in Figure 3, plant 20 is shown as a chemical plant. In plant 20 with the configuration shown in Figure 3, a jacket 22 is provided around the outer circumference of the reaction vessel 21. In other words, the jacket 22 covers the sides and bottom of the reaction vessel 21.

[0025] A raw material supply pipe 23 is connected to the reaction vessel 21. Various raw materials are introduced into the reaction vessel 21 via the raw material supply pipe 23, and the various raw materials undergo chemical reactions inside the reaction vessel 21 to produce the product. At this time, reaction heat is generated by the chemical reaction, causing the temperature of the reaction vessel 21 to fluctuate, but the control goal is to maintain the reaction temperature at a predetermined value in order to maintain the quality of the product.

[0026] The jacket 22 is installed around the outer circumference of the reaction vessel 21 and is designed to lower the temperature of the reaction vessel 21 with cooling water supplied from the cooling water supply pipe 24. The cooling water supply pipe 24 is equipped with a control valve 25 for adjusting the amount of cooling water supplied. The more cooling water supplied through the cooling water supply pipe 24, that is, the greater the opening of the control valve 25, the greater the decrease in the temperature of the reaction vessel 21. The cooling water that has exchanged heat with the reaction vessel 21 via the jacket 22 is discharged from the cooling water discharge pipe 26.

[0027] As shown in Figure 3, the plant 20 is also equipped with a reaction vessel temperature sensor 31, a raw material supply flow rate sensor 32, a control valve opening indicator 33, a cooling water flow rate sensor 34, and a jacket temperature sensor 35. The reaction vessel temperature sensor 31 is installed in the reaction vessel 21, and the temperature T1 of the reaction vessel 21 is monitored by the temperature sensor 31. The raw material supply flow rate sensor 32 is installed in the raw material supply piping 23, and the flow rate F1 of the raw material supplied to the reaction vessel 21 is monitored by the raw material supply flow rate sensor 32. The control valve opening indicator 33 is installed in the control valve 25, and the valve opening CV of the control valve 25 is monitored by the control valve opening indicator 33. The cooling water flow rate sensor 34 is installed in the cooling water supply piping 24, and the flow rate F2 of the cooling water is monitored by the cooling water flow rate sensor 34. The jacket temperature sensor 35 is installed in the jacket 22, and the temperature T2 of the jacket 22 is monitored by the jacket temperature sensor 35. The reaction vessel temperature sensor 31, the raw material supply flow rate sensor 32, the control valve opening indicator 33, the cooling water flow rate sensor 34, and the jacket temperature sensor 35 correspond to the aforementioned sensor 30 (see Figure 1).

[0028] Let's return to Figure 1 and continue the explanation. The data acquisition unit 11 shown in Figure 1 acquires predetermined data from the plant 20, sensor 30, input device 40, cloud 60, and external recording media 70 such as optical disc 70a and USB memory 70b via the communication unit 13. In other words, the data acquisition unit 11 receives predetermined data from the plant 20, sensor 30, input device 40, cloud 60, and external recording media 70 such as optical disc 70a and USB memory 70b via the communication unit 13.

[0029] The memory unit 12 stores various types of information handled by the information processing device 10. As shown in Figure 1, the memory unit 12 stores sensor connection relationship information 12a, sensor measurement data 12b, controlled variable information 12c, manipulated variable information 12d, causal relationship information 12e, influence degree table for controlled variables 12f, influence degree table from manipulated variables 12g, control model input variable selection table 12h, and control model 12i. Details of these will be described later, but in general terms, the following information is stored in the memory unit 12.

[0030] Sensor connection relationship information 12a is information that graphically represents the physical and signal connection relationships of the sensors provided by the plant 20. Sensor connection relationship information 12a can be input by operator M1 via input device 40, or it may be acquired from the cloud 60 or external recording medium 70 via communication unit 13 and data acquisition unit 11 and stored in storage unit 12.

[0031] Figure 4 is an explanatory diagram of sensor connection relationship information in the information processing device shown in Figure 1. In the graph shown in Figure 4, the nodes correspond to the sensors 30 provided by the plant 20, and the edges represent the physical and signal connection relationships of each sensor. For example, the edge between the nodes corresponding to the temperature sensor 31 of the reaction vessel and the flow rate sensor 32 of the raw material supply indicates that the reaction vessel 21 and the raw material supply piping 23, where each sensor is installed, are physically connected. Similarly, other edges indicate that the devices and piping on which the sensors are installed are physically connected.

[0032] Figure 5A is an explanatory diagram showing a specific example of a plant that is the target of the control model generated by the information processing device shown in Figure 1, and is equipped with signal-connected sensors. Figure 5B is an explanatory diagram of the sensor connection relationship information of the plant shown in Figure 5A. In the plant configuration shown in Figure 5A, a fluid supply pipe 201, a tank 202, and a fluid discharge pipe 203 are connected in sequence. The fluid supply pipe 201 is equipped with a control valve 204 for adjusting the amount of fluid supplied to the tank 202. The fluid supplied to the tank 202 is then discharged through the fluid discharge pipe 203.

[0033] Tank 202 is equipped with a liquid level sensor 205, which is connected to the liquid level controller 206 as a measured value PV. The liquid level controller 206 outputs an operating variable MV so that the measured value PV matches the set value SV of the liquid level controller 206, thereby adjusting the opening degree of the control valve 204. The valve opening degree CV of the control valve 204 is monitored by the control valve opening degree indicator 207.

[0034] Figure 5B shows a specific example of sensor connection relationship information 12a for a sensor, including the measured value PV206a, the set value SV206b, and the manipulated variable MV206c of the liquid level controller 206 installed in the plant example in Figure 5A. Directed edges in Figure 5B represent the signal connection of sensors. For example, the edge from liquid level sensor 205L to the measured value PV206a of liquid level controller 206 indicates that liquid level sensor 205 is connected as the measured value PV206a of liquid level controller 206. Also, the edges from the measured value PV206a and the set value SV206b of liquid level controller 206 to the manipulated variable MV206c indicate that the manipulated variable MV is determined from the measured value PV and the set value SV in liquid level controller 206. Undirected edges in Figure 5B, as in Figure 4, indicate that the devices and piping on which the sensors are installed are physically connected.

[0035] Sensor measurement data 12b is information that represents the measured values ​​of each sensor for each date and time in a table format. In addition to being acquired from the sensor 30, the sensor measurement data 12b may also be acquired by operator M1 via the input device 40, or from the cloud 60 or external recording medium 70 via the communication unit 13 and data acquisition unit 11 and stored in the storage unit 12. Figure 6 shows an example of sensor measurement data from the sensors installed in the plant shown in Figure 3. As shown in Figure 6, each row of the sensor measurement data 12b corresponds to measurement data for each date and time, and each column corresponds to measurement data for each sensor 30. For example, when the "Date and Time" is "2024-01-01 00:01:00", the sensor measurement data is as follows: "T1" of the temperature of the reaction vessel 21 measured by the temperature sensor 31 is "67.1", "F1" of the raw material flow rate supplied to the reaction vessel 21 measured by the raw material supply flow rate sensor 32 is "1.1", "T2" of the jacket 22 measured by the jacket temperature sensor 35 is "66.9", "F2" of the cooling water flow rate measured by the cooling water flow rate sensor 34 is "500", and "CV" of the control valve 25 measured by the control valve opening indicator 33 is "0.10".

[0036] The controlled variable information 12c is a list of controlled variables, which are items of the controlled quantity being controlled. For example, in the control of the plant 20 shown in the example in Figure 3, the control objective is to keep the temperature T1 of the reaction vessel 21 constant, so there is only one controlled variable, "T1". In addition to the items that are the control objective, the controlled variables may also include items that are control constraints. The controlled variable information 12c may be acquired from the cloud 60 or an external recording medium 70 and stored in the storage unit 12, in addition to being input by the operator M1 via the input device 40.

[0037] The control variable information 12d is a list of control variables, which are items of the controllable quantities to be controlled. For example, in the control of the plant 20 shown in the example in Figure 3, the control variable is "CV" because the valve opening CV of the control valve 25 is operated in order to adjust the temperature T1 of the reaction vessel 21, which is the variable to be controlled. In addition to being input by the operator M1 via the input device 40, the control variable information 12d may also be acquired from the cloud 60 or external recording medium 70 via the communication unit 13 and the data acquisition unit 11 and stored in the storage unit 12.

[0038] The causal relationship information 12e is information that represents the causal relationships of the measurement items of the sensors installed in the plant 20 as a directed graph. In addition to the calculation results of the causal relationship information generation unit 14a, the causal relationship information 12e may also be acquired from input by operator M1 via the input device 40, or from the cloud 60 or external recording medium 70 via the communication unit 13 and data acquisition unit 11 and stored in the storage unit 12.

[0039] Figure 7 shows an example of causal relationship information for sensors in the plant shown in Figure 3. In the graph of Figure 4, the nodes correspond to the sensors 30 in the plant 20, and the edges represent the physical and signal connection relationships of each sensor. In the graph of Figure 7, the nodes correspond to the sensors 30 in the plant 20, and the directed edges represent the causal relationships of each sensor. For example, the edge from the raw material supply flow rate sensor 32 to the reaction vessel temperature sensor 31 represents a causal relationship where a change in the raw material supply flow rate F1 (cause), which is the parent node, causes a change in the reaction vessel temperature T1 (effect), which is the child node. Also, the double-headed arrow edge between the reaction vessel temperature sensor 31 and the jacket temperature sensor 35 represents a causal relationship in which the reaction vessel temperature T1 and the jacket temperature T2 mutually influence each other.

[0040] The influence table 12f on the controlled variable is information that represents the calculation results of the influence calculation unit 14b on the controlled variable in a table format. "Influence on the controlled variable" refers to the degree to which changes in the measured values ​​of each sensor 30 have an impact on the value of the controlled variable in the controlled variable information 12c. Sensors with a large influence on the controlled variable are useful for predicting the future value of the controlled variable.

[0041] The influence table 12g from the control variable is information that represents the calculation results of the influence calculation unit 14c from the control variable in table format. "Influence from the control variable" refers to the degree to which a change in the value of the control variable in the control variable information 12d has an effect on the value of each sensor 30. A sensor with a high influence from the control variable indicates that its value can be easily controlled by manipulating the control variable.

[0042] The control model input variable selection table 12h is information that represents the calculation results of the control model input variable selection unit 14d in table format. The control model 12i is the result of calculations performed by the control model generation unit 14f, and is a model that predicts the optimal operation of the controlled object using the measured data of the controlled object as input variables. The data format of the control model 12i may be a table, a function, or a predetermined calculation formula.

[0043] The communication unit 13 communicates with the plant 20, sensors 30, input devices 40, display devices 50, cloud 60, and external recording media 70. The communication unit 13 is connected to the storage unit 12 and the arithmetic unit 14 via an internal bus. Data acquired from the plant 20, etc., via the communication unit 13 is stored in the storage unit 12, and calculation results from the arithmetic unit 14 are transmitted to the plant 20, etc., via the communication unit 13.

[0044] The calculation unit 14 generates data related to the control of the plant 20. At least a portion of the processing performed by the calculation unit 14 may be performed by AI. As shown in Figure 1, the calculation unit 14 includes a causal relationship information generation unit 14a, an influence calculation unit 14b for controlled variables, an influence calculation unit 14c for manipulated variables, a control model input variable selection unit 14d, a control model input variable display unit 14e, and a control model generation unit 14f. It is also possible to replace the functions of each functional unit of the calculation unit 14 with the cloud 60. Details of each function of the arithmetic unit 14 will be described later, but in general terms, it has the following functions.

[0045] In other words, the causal relationship information generation unit 14a generates causal relationship information 12e based on the sensor connection relationship information 12a and sensor measurement data 12b stored in the storage unit 12. The causal relationship information 12e generated by the causal relationship information generation unit 14a is stored in the storage unit 12.

[0046] The influence calculation unit 14b calculates an influence table 12f for the controlled variable based on the sensor measurement data 12b, controlled variable information 12c, and causal relationship information 12e stored in the storage unit 12. The influence table 12f for the controlled variable calculated by the influence calculation unit 14b is stored in the storage unit 12.

[0047] The influence calculation unit 14c calculates an influence table 12g from the manipulated variable based on the sensor measurement data 12b, manipulated variable information 12d, and causal relationship information 12e stored in the memory unit 12. The influence table 12g from the manipulated variable calculated by the influence calculation unit 14c is stored in the memory unit 12.

[0048] The control model input variable selection unit 14d calculates a control model input variable selection table 12h based on the controlled variable information 12c, the manipulated variable information 12d, the influence table 12f on the controlled variable, and the influence table 12g from the manipulated variable, all stored in the memory unit 12, and selects the input variables for the control model. By considering both the influence on the controlled variable and the influence from the manipulated variable, it is possible to extract variables that are useful for predicting the future of the controlled variable and are difficult to manipulate with the manipulated variable, thereby selecting input variables suitable for the control model. The control model input variable selection table calculated by the control model input variable selection unit 14d is stored in the memory unit 12.

[0049] The control model input variable display unit 14e displays the control model input variable selection table 12h stored in the memory unit 12 on the display device 50. This allows the operator M1 to easily understand the input variables of the control model of the plant 20.

[0050] The control model generation unit 14f generates a control model 12i based on the sensor measurement data 12b, the operation variable information 12d, and the control model input variable selection table 12h stored in the memory unit 12.

[0051] Furthermore, the following functions are performed: the causal relationship information generation unit 14a generates causal relationship information 12e based on the sensor connection relationship information 12a and sensor measurement data 12b stored in the storage unit 12; the influence degree calculation unit 14b calculates an influence degree table 12f for the controlled variable based on the sensor measurement data 12b, controlled variable information 12c and causal relationship information 12e stored in the storage unit 12; the influence degree calculation unit 14c calculates an influence degree table 12g from the manipulated variable based on the sensor measurement data 12b, manipulated variable information 12d and causal relationship information 12e stored in the storage unit 12; and the control model input variable selection unit 14d selects the controlled variable information 12c and manipulated variable information stored in the storage unit 12. The functions of the control model input variable selection table 12h, which calculates the control model input variable selection table 12h based on the manipulated variable information 12d, the influence degree table 12f on the controlled variable, and the influence degree table 12g from the manipulated variable, and then selecting the input variables for the control model; the control model input variable display unit 14e, which displays the control model input variable selection table 12h stored in the storage unit 12 on the display device 50; and the control model generation unit 14f, which generates the control model 12i based on the sensor measurement data 12b, the manipulated variable information 12d, and the control model input variable selection table 12h stored in the storage unit 12, are each stored in the ROM 10c and storage 10d as programs executed by the processor 10a (see Figure 2).

[0052] <Processing by information processing device> Figure 8 is a flowchart showing the processing flow of the information processing device shown in Figure 1 (refer to Figure 1 as needed). The series of processes shown in Figure 8 are initiated by the operator M1 via the input device 40. In step S1 of Figure 8, the data acquisition unit 11, which constitutes the information processing device 10, acquires the input information necessary for processing by the information processing device 10 and stores it in the storage unit 12. Specifically, sensor connection relationship information 12a, sensor measurement data 12b, controlled variable information 12c, and operation variable information 12d are acquired via the communication unit 13 from inputs from the plant 20, sensors 30, and operator M1 via input devices 40, as well as from the cloud 60 and external recording medium 70, and stored in the storage unit 12.

[0053] Next, in step S2, the causal relationship information generation unit 14a, which constitutes the calculation unit 14 of the information processing device 10, generates causal relationship information 12e based on the sensor connection relationship information 12a and sensor measurement data 12b stored in the storage unit 12. For the specific processing in step S2, predetermined causal search methods such as PCMCI (Peter-Clark and Momentary Conditional Independence) and VAR-LiNGAM (Vector Auto Regression-Linear Non-Gaussian Acyclic Model) can be used. This estimates the direction of the causal relationship of the undirected edges in the graph of the sensor connection relationship information 12a from the data and generates the causal relationship information 12e.

[0054] Next, in step S3, the influence degree calculation unit 14b of the information processing device 10, which constitutes the calculation unit 14, calculates the influence degree table 12f of the controlled variable based on the sensor measurement data 12b, the controlled variable information 12c, and the causal relationship information 12e, and stores it in the storage unit 12. The specific method for calculating the influence degree will be described later, but in step S3, the influence degree table 12f of the controlled variable is calculated by calculating the influence degree of the controlled variable from the sensor item for each sensor item.

[0055] Next, in step S4, the influence calculation unit 14c from the manipulated variables, which constitutes the calculation unit 14 of the information processing device 10, calculates the influence table 12g from the manipulated variables based on the sensor measurement data 12b, manipulated variable information 12d, and causal relationship information 12e, and stores it in the storage unit 12. The specific method for calculating the influence will be described later, but in step S4, the influence table 12g from the manipulated variables is calculated for each sensor item by calculating the influence of the manipulated variables on the sensor item.

[0056] In this embodiment, we show an example where steps S3 and S4 are processed in parallel, but this is not the only way. For example, step S4 may be processed after step S3, or step S3 may be processed after step S4.

[0057] After the processing in steps S3 and S4 is completed, in step S5 of Figure 8, the control model input variable selection unit 14d, which constitutes the calculation unit 14 of the information processing device 10, generates a control model input variable selection table 12h based on the controlled variable information 12c, the manipulated variable information 12d, the influence degree table 12f on the controlled variable, and the influence degree table 12g from the manipulated variable, and stores it in the storage unit 12. The specific processing in step S5 will be described later, but for each sensor item, an evaluation value is calculated based on the influence degree on the controlled variable and the influence degree from the manipulated variable, and based on that evaluation value, it is determined whether or not to select it as an input variable of the control model, and the control model input variable selection table 12h is generated.

[0058] Next, in step S6, the control model input variable display unit 14e, which constitutes the calculation unit 14 of the information processing device 10, displays the control model input variable selection table 12h on the display device 50.

[0059] Next, in step S7, the control model generation unit 14f, which constitutes the calculation unit 14 of the information processing device 10, generates a control model 12i based on the sensor measurement data 12b, the manipulated variable information 12d, and the control model input variable selection table 12h, and stores it in the storage unit 12. Specifically, it extracts the manipulated variables from the manipulated variable information 12d and the sensor items selected in the control model input variable selection table 12h from the data items of the sensor measurement data 12b to create training data, and generates the control model 12i using a predetermined machine learning technique. Here, conservative Q-learning, one of the offline reinforcement learning techniques, was used to generate the control model. Conservative Q-learning is effective in safely controlling a controlled object because it reduces risk by not overestimating the value of each operation of the controlled object.

[0060] <Calculation of the degree of influence on the controlled variable and the degree of influence from the manipulated variable> Figure 9A is an explanatory diagram illustrating the degree of influence on the controlled variable in the plant shown in Figure 3. Figure 9B is an explanatory diagram illustrating the degree of influence from the manipulated variable in the plant shown in Figure 3. Each node in Figures 9A and 9B corresponds to each sensor, and the degree to which a change in the value of the parent node of the edge indicated by the thick arrow affects the value of the child node represents the degree of influence on the controlled variable (Figure 9A) and the degree of influence from the manipulated variable (Figure 9B). In the example shown in Figures 9A and 9B, the controlled variable and the manipulated variable are the temperature T1 of the reaction vessel 21 and the valve opening CV of the control valve 25, respectively.

[0061] Figure 10 shows an example of an influence table for the controlled variable in the plant shown in Figure 3. Figure 10 also shows an example of the result of calculating the influence on the controlled variable, which is the process in step S3 of Figure 8. Specifically, in step S3, the influence calculation unit 14b of the controlled variable, which constitutes the calculation unit 14 of the information processing device 10, calculates the influence δ from each sensor item X to the controlled variable S based on the following equation (1). X→S Figure 10 shows an example of the influence table 12f on the controlled variable, which was generated by calculating the influence of the variable.

[0062]

number

[0063] Note that E[S|do(X=x)] in equation (1) is the expected value of the controlled variable S when the sensor item X is set to x. Also, μ in equation (1) X Ya σ X Ya σ S These are the mean and standard deviation of sensor item X and the standard deviation of the controlled variable S, respectively, and |·| represents the absolute value. That is, the degree of influence on the controlled variable δ calculated by equation (1) X→S This represents how many standard deviations the value of the controlled variable S changes by when the value of sensor item X is varied by one standard deviation around the mean.

[0064] The value of E[S|do(X=x)] in equation (1) can be calculated by the following process. First, based on the causal relationship information 12e, the confounding factor C, which is the parent node of both the sensor item X and the controlled variable S, is identified. Next, based on the sensor measurement data 12b, the expected value of the controlled variable S E[S|X=x,C=c] and the probability distribution P(C) of the confounding factor C are calculated when conditioned by the values ​​of the sensor item X and the confounding factor C. Then, based on the following equation (2), the value of E[S|do(X=x)] in equation (1) is calculated.

[0065]

number

[0066] As described above, the influence table 12f on the controlled variable in Figure 10 is generated by calculating the influence on the controlled variable for each sensor item based on equations (1) and (2), sensor measurement data 12b, and causal relationship information 12e. The influence from one controlled variable to another is defined as 1.

[0067] Figure 11 shows an example of an influence table from the operational variables in the plant shown in Figure 3. Figure 11 also shows an example of the result of calculating the influence from the operational variables, which is the process in step S4 of Figure 8. Specifically, in step S4, the influence calculation unit 14c from the operational variables, which constitutes the calculation unit 14 of the information processing device 10, calculates the influence δ from operational variable A to each sensor item X based on the following equation (3). A→X Figure 11 shows an example of an influence table 12g generated from the instrumental variables by calculating the following:

[0068]

number

[0069] Note that E[X|do(A=a)] in equation (3) is the expected value of sensor item X when the manipulated variable A is set to a, and can be calculated in the same way as in equation (2). Also, μ in equation (3) A Ya σA and σ X are the average value and standard deviation of manipulated variable A, and the standard deviation of sensor item X, respectively, and |·| represents an absolute value. That is, the influence degree δ from the manipulated variable calculated by equation (3) A→X represents how many standard deviations the value of sensor item X changes when the value of manipulated variable A is changed by one standard deviation around the average value.

[0070] As described above, the influence degree table 12g from manipulated variables in FIG. 11 is generated by calculating the influence degree from manipulated variables for each sensor item based on equations (3), equation (2), sensor measurement data 12b, and causal relationship information 12e. Note that the influence degree from a manipulated variable to itself is 1 by definition.

[0071] <Selection of input variables for control model> FIG. 12 is a diagram showing an example of a control model input variable selection table in the plant shown in FIG. 3. FIG. 12 shows an example of a result of selection of input variables for a control model, which is the processing of step S5 in FIG. 8. That is, in step S5, the control model input variable display unit 14e constituting the calculation unit 14 of the information processing apparatus 10 calculates evaluation values for each sensor item based on the influence degree table 12f to controlled variables and the influence degree table 12g from manipulated variables, and the table in FIG. 12 is an example of a result of selecting input variables for a control model based on the evaluation values.

[0072] The evaluation value V of each sensor item X in FIG. 12 X is calculated using the following equation (4) based on the influence degree δ of each sensor item X on the controlled variable X→S and the influence degree δ from the manipulated variable A→X Note that calculation of the evaluation value can be omitted for controlled variables and manipulated variables.

[0073] [Mathematics]

[0074] Then, the evaluation value V of each sensor item X XBy determining whether the value is greater than a predetermined threshold, it is determined whether or not to select it as an input variable for the control model. Figure 12 shows an example where the threshold is set to 0, indicating that sensor items in the Control Model Input Variables column that are "TRUE" are selected as input variables for the control model, while sensor items that are "FALSE" are not selected.

[0075] <Displaying input variables of the control model> In step S6 of Figure 8, the control model input variable display unit 14e, which constitutes the calculation unit 14 of the information processing device 10, displays the control model input variable selection table 12h shown in Figure 12 on the display device 50. In addition, it can further display which sensor items were selected as input variables for the control model, in comparison with the causal relationships of each sensor, using causal relationship information 12e, the influence degree table 12f on the controlled variable, and the influence degree table 12g from the manipulated variable.

[0076] Figure 13 shows an example of displaying the sensors selected as input variables for the control model in the plant shown in Figure 3, overlaid on the plant configuration diagram. Each node in Figure 13 corresponds to a sensor, with controlled variables enclosed in squares (□), manipulated variables in diamonds (◇), and input variables of the control model other than controlled variables in circles (○). The edges indicated by solid and dashed arrows represent the degree of influence on the controlled variable and the degree of influence from the manipulated variable, respectively, through the thickness and intensity of the arrows. This allows operator M1 to easily understand the position of the sensors selected as input variables for the control model 12i on the plant 20, as well as their relationship to the controlled and manipulated variables, which can be used to evaluate the reliability of the control model 12i.

[0077] Figure 14 is an example of displaying the degree of influence of each sensor on the controlled variable and the degree of influence from the control variable in the plant shown in Figure 3 on a two-dimensional plane. In Figure 14, the horizontal axis represents the degree of influence of each sensor on the controlled variable, and the vertical axis represents the degree of influence of each sensor from the control variable. Each point in Figure 14 corresponds to each sensor. The dashed line in Figure 14 represents the threshold value of the evaluation value calculated by equation (4), and the shaded area indicates that the evaluation value is below the threshold. In other words, sensor items in the shaded area are not selected as input variables for the control model 12i. Sensor items selected as input variables for the control model 12i are represented by ★. This allows operator M1 to easily understand the criteria for selecting input variables for the control model and use this to evaluate the reliability of the control model 12i.

[0078] <Effects> In this embodiment, the degree of influence of each sensor item on the controlled variable (Figure 9A) and the degree of influence from the manipulated variable (Figure 9B) are used to select the input variables for the control model 12i and generate the control model 12i. This makes it possible to extract variables that are useful for predicting the future of the controlled variable and that are difficult to manipulate with the manipulated variable, and to select input variables suitable for the control model 12i to generate a control model 12i with good controllability.

[0079] Furthermore, equations (1) and (3) are used to calculate the degree of influence on the controlled variable and the degree of influence from the instrumental variable. This allows the degree of influence to be calculated based on the standard deviation of the sensor item values, making it possible to compare the degree of influence even for sensor items with different data scales by unifying the scale.

[0080] Furthermore, the sensor items selected as input variables for the control model 12i are displayed overlaid on the plant configuration diagram (Figure 13), and the degree of influence on the controlled variable and the degree of influence from the manipulated variable are displayed on a two-dimensional plane (Figure 14). This allows operator M1 to easily understand the criteria for selecting input variables for the control model 12i, which can be used to evaluate the reliability of the control model 12i.

[0081] <Variation> Although the information processing device 10, etc., according to Example 1 have been described above, the present invention is not limited to these descriptions and various modifications can be made. For example, while Example 1 described a case where there is only one controlled object, Plant 20, Example 1 is applicable not only to cases where there is one controlled object, but also to cases where there are multiple controlled objects.

[0082] Furthermore, while Embodiment 1 described a configuration in which the information processing device 10 (see Figure 1) includes a communication unit 13 (see Figure 1), the configuration is not limited to this. That is, the communication unit 13 may be omitted from the configuration in Figure 1 as appropriate, and the control model input variable display unit 14e may directly transmit the display screen to the display device 50.

[0083] Alternatively, the influence table 12f (see Figure 10) on the controlled variable, the influence table 12g (see Figure 11) from the manipulated variable, and the control model input variable selection table 12h (see Figure 12) may be displayed on the display screen of the display device 50.

[0084] Furthermore, while Example 1 described the case where the influence on the controlled variable calculation unit 14b and the influence on the manipulated variable calculation unit 14c are calculated based on equations (1) and (3), the method is not limited to this. For example, the upper limit value x of sensor item X max and lower limit x min and the upper limit s of the controlled variable S max and lower limit s min Alternatively, the degree of influence on the controlled variable may be calculated using the following equation (5).

[0085]

number

[0086] The degree of influence δ on the controlled variable S calculated by equation (5) X→SThis represents how many units of the upper and lower limits the value of the controlled variable S changes when the value of sensor item X is varied from the lower limit to the upper limit. This allows for the appropriate calculation of the degree of influence even when the distribution of sensor item values ​​deviates from a normal distribution. Note that the degree of influence from the instrumental variable δ A→X The same formula as in formula (5) may also be used to calculate this.

[0087] Furthermore, while Example 1 described a case in which the control model input variable selection unit 14d calculates the evaluation value of each sensor item in the control model input variable selection table 12h (Figure 12) based on equation (4), it is not limited to this. For example, the degree of influence δ on the controlled variable X→S and the influence of the instrumental variable δ A→X Weighted differences for each of these, or at least the influence δ on the controlled variable. X→S and the influence of the instrumental variable δ A→X You may also use various calculation formulas that take the input as the input.

[0088] Furthermore, the aforementioned program can be provided via communication lines, or it can be written to a recording medium such as a CD-ROM and distributed.

[0089] As described above, this embodiment makes it possible to provide an information processing device, an information processing method, and an information processing program that can appropriately generate a control model that predicts the optimal operation of a controlled object.

[0090] It should be noted that the present invention is not limited to the embodiments described above, and various modifications are included. For example, the embodiments described above are explained in detail to make the present invention easier to understand, and are not necessarily limited to those having all the configurations described.

[0091] Furthermore, some or all of the above-mentioned configurations, functions, processing units, and processing methods may be implemented in hardware, for example, by designing them as integrated circuits. Alternatively, the above-mentioned configurations and functions may be implemented in software by having the processor interpret and execute programs that implement each function. Information such as programs, tables, and files that implement each function can be stored in memory, a recording device such as a hard disk or SSD (Solid State Drive), or a recording medium such as an IC card, SD card, or DVD.

[0092] Furthermore, the control lines and information lines shown are those deemed necessary for explanatory purposes, and not all control lines and information lines are necessarily shown in the actual product. In reality, it is reasonable to assume that almost all components are interconnected. [Explanation of Symbols]

[0093] 10…Information Processing Devices 10a…Processor 10b…RAM 10c…ROM 10d...Storage 10e…Communication Interface 10g…I / O I / F 10h…Media I / F 10i...Internal bus 11...Data acquisition unit 12...Storage section 12a... Sensor connection information 12b...Sensor measurement data 12c... Information on controlled variables 12d...Information about the manipulated variable 12e…Causal relationship information 12f...Table of influence on the controlled variable 12g...Influence Table from Instrumental Variables 12h... Control Model Input Variable Selection Table 12i…Control Model 12i 13… Communications Department 14...Arithmetic section 14a...Causal relationship information generation unit 14b...Influence calculation unit for the controlled variable 14c...Influence calculation unit from the manipulated variable 14d...Control model input variable selection section 14e...Control model input variable display section 14f...Control model generation unit 20...Plant 21…Reaction vessel 22…Jacket 23...Raw material supply piping 24…Cooling water supply piping 25,204... Control valve 26…Cooling water discharge piping 30...Sensor 31…Temperature sensor for the reaction vessel 32…Flow sensor for raw material supply 33,207... Control valve opening indicator 34…Cooling water flow sensor 35… Jacket temperature sensor 40…Input device 50…Display device 60…Cloud 70…External recording media 70a…Optical disc 70b... USB memory 100… Information Processing Systems 201...Fluid supply piping 202... Tank 203…Fluid discharge piping 205... Liquid level sensor 206…Liquid level controller

Claims

1. A data acquisition unit that accepts input of sensor connection relationship information, sensor measurement data, operating variable information of the controlled object, and controlled variable information of the controlled object. A causal relationship information generation unit generates causal relationship information based on the aforementioned sensor connection relationship information and sensor measurement data, An impact on a controlled variable calculation unit calculates the degree of impact on the controlled variable, which indicates the degree to which a change in the sensor measurement data affects the controlled variable, based on the sensor measurement data, controlled variable information, and causal relationship information. An influence calculation unit from the control variable calculates the degree of influence from the control variable, which indicates the degree to which a change in the control variable affects the sensor measurement data, based on the sensor measurement data, control variable information, and causal relationship information. An information processing apparatus characterized by comprising a selection unit that selects input variables for a control model based on the degree of influence on the controlled variable and the degree of influence from the manipulated variable.

2. In the information processing apparatus according to claim 1, An information processing device characterized by having a display unit that displays the input variables of the control model selected by the selection unit in correspondence with the evaluation values ​​of the sensor items.

3. In the information processing apparatus according to claim 2, The information processing device is characterized in that the display unit displays the degree of influence of the input variables and controlled variables of the control model, as well as the degree of influence from the manipulated variables, overlaid on the causal relationship information.

4. In the information processing apparatus according to claim 2, The information processing device is characterized in that the display unit displays the input variables of the control model on a two-dimensional plane with axes representing the degree of influence from the manipulated variable and the degree of influence on the controlled variable.

5. In the information processing apparatus described in claim 3, The display unit is characterized in that it represents the degree of influence on the controlled variable and the degree of influence from the manipulated variable by the thickness or intensity of an arrow.

6. In the information processing apparatus described in claim 3, The unit for calculating the degree of influence on the controlled variable calculates the degree of influence on the controlled variable based on the standard deviation of each sensor and the controlled variable. The information processing device is characterized in that the influence calculation unit from the control variable calculates the influence from the control variable based on the standard deviation of each sensor.

7. In the information processing apparatus according to claim 4, The unit for calculating the degree of influence on the controlled variable calculates the degree of influence on the controlled variable based on the standard deviation of each sensor and the controlled variable. The information processing device is characterized in that the influence calculation unit from the control variable calculates the influence from the control variable based on the standard deviation of each sensor.

8. An information processing method for an information processing apparatus comprising: a data acquisition unit; a causal relationship information generation unit; an influence calculation unit for a controlled variable; an influence calculation unit from an manipulated variable; and a selection unit; The data acquisition unit receives input of sensor connection relationship information, sensor measurement data, operating variable information of the controlled object, and controlled variable information of the controlled object. The aforementioned causal relationship information generation unit generates causal relationship information based on the sensor connection relationship information and sensor measurement data. The unit that calculates the degree of influence on the controlled variable calculates the degree of influence on the controlled variable, which indicates the degree of influence on the controlled variable due to a change in the sensor measurement data, based on the sensor measurement data, the controlled variable information, and the causal relationship information. The influence calculation unit from the control variable calculates the degree of influence from the control variable, which indicates the degree to which the change in the control variable affects the sensor measurement data, based on the sensor measurement data, control variable information, and causal relationship information. The information processing method is characterized in that the selection unit selects input variables for a control model based on the degree of influence on the controlled variable and the degree of influence from the manipulated variable.

9. In the information processing method described in claim 8, An information processing method characterized in that the display unit displays the input variables of the control model selected by the selection unit in correspondence with the evaluation values ​​of the sensor items.

10. In the information processing method described in claim 9, An information processing method characterized in that the display unit overlays the degree of influence of the input variables and controlled variables of the control model, as well as the degree of influence from the manipulated variables, onto the causal relationship information.

11. In the information processing method described in claim 9, The information processing method is characterized in that the display unit displays the input variables of the control model on a two-dimensional plane with axes representing the degree of influence from the manipulated variable and the degree of influence on the controlled variable.

12. In the information processing method described in claim 10, An information processing method characterized in that the display unit represents the degree of influence on the controlled variable and the degree of influence from the manipulated variable by the thickness or intensity of an arrow.

13. In the information processing method described in claim 10, The unit that calculates the degree of influence on the controlled variable calculates the degree of influence on the controlled variable based on the standard deviation of each sensor and the controlled variable. An information processing method characterized in that the influence calculation unit from the control variable calculates the influence from the control variable based on the standard deviation of each sensor.

14. In the information processing method described in claim 11, The unit that calculates the degree of influence on the controlled variable calculates the degree of influence on the controlled variable based on the standard deviation of each sensor and the controlled variable. An information processing method characterized in that the influence calculation unit from the control variable calculates the influence from the control variable based on the standard deviation of each sensor.

15. In the processor, A function that accepts input of sensor connection relationship information, sensor measurement data, operating variable information of the controlled object, and controlled variable information of the controlled object. A function to generate causal relationship information based on the aforementioned sensor connection relationship information and sensor measurement data, Based on the aforementioned sensor measurement data, controlled variable information, and causal relationship information, a function is provided to calculate the degree of impact on the controlled variable, which indicates the degree to which a change in the sensor measurement data affects the controlled variable. Based on the aforementioned sensor measurement data, control variable information, and causal relationship information, a function is provided to calculate the degree of influence from the control variable, indicating the degree to which a change in the control variable affects the sensor measurement data. An information processing program for implementing a function to select input variables for a control model based on the degree of influence on the controlled variable and the degree of influence from the manipulated variable.

Images